Instant haze removal from a single image

Images captured in bad weather are usually be degraded compared to those in clear days. For this, we draw on atmospheric optics and we propose a simple but powerful approach to remove haze from a single image based on imaging model under atmospheric scattering. For images with nearby objects under dense fog or distant objects under mild fog, we estimate the scene structure as transmission from the image model based on the optical irradiance characteristic which depends solely on its depth. Moreover, we present a reliable statistical principle for estimating global atmospheric light. Incorporating these two estimations into the haze imaging model, we can recover the contrast and color quality as well as that of a haze-free image. Experimental results on a variety of hazy images have shown the power of the methodology.